CN107146013A - A kind of classifying type electric automobile demand spatial and temporal distributions dynamic prediction method based on gray prediction and SVMs - Google Patents
A kind of classifying type electric automobile demand spatial and temporal distributions dynamic prediction method based on gray prediction and SVMs Download PDFInfo
- Publication number
- CN107146013A CN107146013A CN201710297707.7A CN201710297707A CN107146013A CN 107146013 A CN107146013 A CN 107146013A CN 201710297707 A CN201710297707 A CN 201710297707A CN 107146013 A CN107146013 A CN 107146013A
- Authority
- CN
- China
- Prior art keywords
- mrow
- electric automobile
- msub
- data
- prediction
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 91
- 238000012706 support-vector machine Methods 0.000 title claims abstract description 43
- 238000009826 distribution Methods 0.000 title claims description 41
- 230000002123 temporal effect Effects 0.000 title claims description 36
- 238000006467 substitution reaction Methods 0.000 claims abstract description 50
- 238000011161 development Methods 0.000 claims abstract description 15
- 238000011160 research Methods 0.000 claims abstract description 12
- 230000006870 function Effects 0.000 claims description 29
- 238000012545 processing Methods 0.000 claims description 24
- 238000004458 analytical method Methods 0.000 claims description 19
- 230000008569 process Effects 0.000 claims description 18
- 238000009499 grossing Methods 0.000 claims description 14
- 238000012549 training Methods 0.000 claims description 12
- 238000012360 testing method Methods 0.000 claims description 8
- 230000007704 transition Effects 0.000 claims description 7
- 230000015572 biosynthetic process Effects 0.000 claims description 6
- 230000006872 improvement Effects 0.000 claims description 6
- 239000011159 matrix material Substances 0.000 claims description 6
- 230000001373 regressive effect Effects 0.000 claims description 6
- 238000012546 transfer Methods 0.000 claims description 6
- 230000001186 cumulative effect Effects 0.000 claims description 5
- 238000000605 extraction Methods 0.000 claims description 4
- 238000011835 investigation Methods 0.000 claims description 4
- 238000010606 normalization Methods 0.000 claims description 4
- 241000170545 Iago Species 0.000 claims description 3
- 238000000342 Monte Carlo simulation Methods 0.000 claims description 3
- 238000003672 processing method Methods 0.000 claims description 3
- 239000004744 fabric Substances 0.000 claims 1
- 230000008901 benefit Effects 0.000 abstract description 5
- 238000010219 correlation analysis Methods 0.000 abstract description 2
- 239000000523 sample Substances 0.000 description 27
- 230000018109 developmental process Effects 0.000 description 12
- 238000010586 diagram Methods 0.000 description 7
- 230000006399 behavior Effects 0.000 description 5
- 230000005611 electricity Effects 0.000 description 4
- 238000010276 construction Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 238000012417 linear regression Methods 0.000 description 2
- 238000004904 shortening Methods 0.000 description 2
- 208000019901 Anxiety disease Diseases 0.000 description 1
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 1
- 230000036506 anxiety Effects 0.000 description 1
- 230000003542 behavioural effect Effects 0.000 description 1
- 229910052799 carbon Inorganic materials 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 230000002950 deficient Effects 0.000 description 1
- 230000006866 deterioration Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 239000002360 explosive Substances 0.000 description 1
- 230000027648 face development Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000007935 neutral effect Effects 0.000 description 1
- 238000011017 operating method Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000003014 reinforcing effect Effects 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
- G06Q10/06315—Needs-based resource requirements planning or analysis
-
- G06Q50/40—
Abstract
The present invention is applied to field of power, specifically related to a kind of classifying type electric automobile demand dynamic prediction method based on gray prediction and SVMs.Including：First with the fixed updated gray correlation analysis of high precision, different type vehicle fleet size is predicted；The non-linear behavior of different type electric automobile proportion and influence factor is next based on, using Support vector regression method, classifying type electric automobile substitution ratio is obtained using forecast sample and predicts the outcome, constantly corrected and predicted the outcome using iterative method；Finally the above two are predicted the outcome and matched according to car category, classifying type electric automobile demand growth forecast model is set up, with reference to user's trip law study, accurate dynamic space-time prediction is realized to classifying type electric automobile demand.Therefore the invention has the advantages that：The influence of the characteristics of historical data is not enough and different factors to Development of Electric Vehicles is taken into full account, more accurate dynamic prediction is realized in the research for rule of being gone on a journey with reference to user.
Description
Technical field
The present invention is applied to field of power, and it is accurately reasonable to be carried out for the quantity demand to classifying type electric automobile
Prediction, in order to the accurate analysis and the construction of making rational planning for of charging station of the charging load of electric automobile, and in particular to a kind of
Classifying type electric automobile demand dynamic prediction method based on gray prediction and SVMs.
Background technology
Energy industry is the basic industry of national economy, is to ensure that the prerequisite of national strategy safety, while being also real
The important leverage of existing sustainable economic development.However as the continuous expansion of China's economic scale, China is to traditional energy such as oil
The demand in source not only increases, and the carbon content being discharged into air is also more and more so that energy scarcity and the deterioration of the ecological environment
Pressure increasingly increases.It is serious as solution energy resources anxiety, atmosphere pollution and electric automobile is with the advantage of its energy-saving and environmental protection
Effective way.
Ev industry has been classified as the seven great strategy new industries of China, and the Chinese government releases a series of encouragement and helped
Policy is held, accelerates to promote ev industry development.Electric automobile recoverable amount rapid growth, by by the end of August, 2015, China's electricity
Electrical automobile recoverable amount reaches 22.30 ten thousand, and country clearly proposes to reach 5,000,000 to the year two thousand twenty electric automobile recoverable amount.
With the science and technology such as the growth of battery life, the increase of battery capacity and the shortening in electrically-charging equipment charging interval
Progressive, while the guiding energetically of national policy and government widely popularize, electric automobile quantity will show explosive increase
Trend, this will bring huge challenge to the construction scale and service ability of electrically-charging equipment.In order to realize electric automobile
Charging infrastructure and the target of the synchronized development of electric automobile scale, it is necessary to which the demand growth to electric automobile is made accurately
Prediction, therefore it is highly desirable to explore the accurate method of classifying type electric automobile demand dynamic prediction.
Current prediction of the domestic scholars to electric automobile recoverable amount has carried out substantial amounts of research, but Consideration is more single
One, also seldom carry out classification and probe into, mainly estimated using single methods such as linear regressions, this is protected to classifying type electric automobile
The forecasting accuracy for the amount of having has a certain impact, and for classifying type electric automobile demand spatial and temporal distributions dynamic prediction in terms of
Research is still in blank.
The content of the invention
Because scholar considers that influence factor is single, accuracy is poor, no for the prediction of electric automobile recoverable amount before
With limitations such as dynamics, this paper presents combine grey method and SVMs based on updating substitution ratio model
Classifying type electric automobile demand dynamic prediction method, accurately dynamic is carried out to the recoverable amount of classifying type electric automobile pre-
Survey, on this basis, with reference to user's trip law study, the spatial and temporal distributions to electronic private car and electric taxi are carried out
Prediction, with reference to the other types electric automobile with fixed driving path feature, realizes electric automobile demand spatial and temporal distributions dynamic
Prediction.First according to the traveling feature and function of electric automobile, electric automobile can be divided into electric bus, electronic taxi
Car, electronic special-purpose vehicle, electronic officer's car and the private class of passenger car four, and probe into the behavior mould of different types of electric automobile
Formula, Accurate Prediction is carried out using improved grey model forecast model to classifying type automobile quantity；Then battery life growth, battery are analyzed
The influences of the technological progress factor to different type electric automobile substitution ratio such as capacity increase, the shortening of electrically-charging equipment charging interval,
And guiding of the national policy to different type electric automobile, SVMs iteration forecast model is built, it is electronic to classifying type
Automobile substitution ratio is predicted；Then combining classification type automobile quantitative forecast result and classifying type electric automobile substitution ratio
Predict the outcome, component updates substitution ratio model, realize the dynamic prediction to classifying type electric automobile demand；Finally, based on pair
The research of user behavior rule, combining classification type electric automobile demand predicts the outcome, to electronic private car and electronic taxi
The spatial and temporal distributions characteristic of car is predicted, realizes classifying type electric automobile demand spatial and temporal distributions dynamic prediction.
The invention belongs to technical field of power systems, be a kind of combination grey method and SVMs is set up based on
The classifying type electric automobile demand dynamic prediction of substitution ratio model is updated, and combines user and goes on a journey rule to electric automobile demand
The method that spatial and temporal distributions are predicted.Its step is as follows：
A kind of classifying type electric automobile demand dynamic prediction method based on gray prediction and SVMs, its feature exists
In, including：
Step 1, classifying type vehicle fleet size forecast model is set up, by region bus, taxi, special-purpose vehicle, public affair
The data characteristicses analysis of the different type vehicle such as car and private passenger car, selection grey method is predicted；To traditional grey
Predicted method carries out research and is improved for its limitation and different type vehicle data feature, obtains the improvement ash of high accuracy
Color forecast model, builds classifying type vehicle fleet size forecast model；By the development trend of the region different type vehicle obtained by investigation,
By classifying type vehicle fleet size forecast model, following overall development trend of the region different type vehicle is predicted；
Step 2, classifying type electric automobile substitution ratio iteration forecast model is set up, by region electric bus, electricity
The ratios of the type vehicle shared by different type vehicle such as dynamic taxi, electronic special-purpose vehicle, electronic officer's car and private passenger car
Data data characteristicses analysis, and influence factor non-linear behavior, selection Support vector regression method obtain decision-making recurrence
Equation, using the substitution ratio data prediction result to classifying type electric automobile of forecast sample, is constantly corrected using iterative method
Predict the outcome, build the electric automobile classifying type substitution ratio iteration forecast model based on SVM；
Step 3, classifying type electric automobile demand growth forecast model is set up；Combining classification classifying type vehicle fleet size predicts mould
Type predicts the outcome to be predicted the outcome with classifying type electric automobile substitution ratio iteration forecast model, and both are predicted the outcome according to automobile
Type is matched, and sets up classifying type electric automobile demand growth forecast model, real to classifying type electric automobile quantity in region
Existing accurate dynamic prediction.
In a kind of above-mentioned classifying type electric automobile demand dynamic prediction method based on gray prediction and SVMs,
In the step 1, improved grey model forecast model is set up, including：
Step 2.1, initial data smoothing step, are improved generation formula as follows：
Original data sequence two ends data smoothing processing mode：
x(1)(1)=(3x(0)(1)+x(0)(2))/4
x(1)(n)=(x(0)(n-1)+3x(0)(n))/4
Original data sequence intermediate data smoothing processing mode：
x(1)(m)=(x(0)(m-1)+2x(0)(m)+x(0)(m+1))/4 1 ＜ m ＜ n
, instead of traditional Accumulating generation mode, more smooth data can be obtained using above processing method；
Step 2.2, Grey Differential Equation is set up, based on equation below：
Model parameter vector α is obtained with least square method i.e.Obtain GM (1,1) model；
Step 2.3, using improving rear backdrop value, the step of setting up difference equation, for the back of the body of the difference equation of high growth
Scape value is modified；
For cumulative data exponential form, i.e. x can be processed data into the situation of high growth trend(1)(t)=μ eat,
The formula is brought into integral formula, i.e.,It can obtain
For it is flat increase or it is incremental it is long using close to formation sequence as background value, i.e.,：
z(1)(k)=0.5x(1)(k)+0.5x(1)(k-1)
Step 2.4, the step of tested to model：The model having built up is tested, examined using test set
Model is suitable, if suitable feasible, carries out next step；If being unsatisfactory for error requirements, the first step is returned to, to data
A smoothing processing is carried out again and otherwise returns to the first step, and data are carried out with a smoothing processing again using formula；
Step 2.5, inverse growth are predicted the outcome：The result obtained is reduced, inverse growth process is carried out, that is,
That is inverse accumulated generating IAGO, will carry out according to the number of times of Accumulating generation, really be predicted the outcome herein.
In a kind of above-mentioned classifying type electric automobile demand dynamic prediction method based on gray prediction and SVMs,
In the step 1, the foundation of classifying type vehicle fleet size forecast model includes：
Step 3.1, initial data is handled：It regard the historical data of the vehicle of a certain type in recent years as original number
According to using the Accumulating generation method after improvement, being smoothed to initial data；
Step 3.2, the suitable background value of selection, difference equation is accurately converted into by the differential equation：According to the first step to original
The processing of beginning data, is analyzed data substantially growth trend, if initial data increases or low growth to be flat, utilizes original
Beginning method using background value is replaced close to formation sequence, is further calculated；If initial data is high growth trend, using changing
The background value generation formula entered, does further calculating after bringing into；
Step 3.3, using least square method, obtain parameter vector α to be asked；
Step 3.4, the model having built up is tested, judge whether the solution of the differential equation is suitable, if properly, entering
Row next step, otherwise returns to the first step, to the further processing of data；
Step 3.5, according to the suitable difference equation of foundation result is predicted, i.e., difference equation solved；
Step 3.6, the result obtained is reduced, inverse growth process, that is, inverse accumulated generating are carried out, according to cumulative life
Into number of times carry out, obtain predicting the outcome for certain final type of vehicle quantity.
In a kind of above-mentioned classifying type electric automobile demand dynamic prediction method based on gray prediction and SVMs,
In the step 2, in the SVMs iteration forecast model, input vector x ∈ RmFor electric automobile substitution ratio history
Data, technological progress influence factor, government policy guiding etc., output y are renewal substitution ratio predicted value；According to given data
Training sample set and forecast sample collection are set up, SVM regressive object functions are set up, optimal solution is solved and brings back to recurrence decision function side
Journey, obtains regression forecasting function, finally calculates and predicts the outcome, specifically includes：
Step 4.1, historical data is pre-processed and normalized；
WhereinFor the data value after normalization, xiFor actual, historical data, xi min=min (xi), xi max=max (xi),
M is the dimension of input vector, that is, influences the number of predictive factorses；
Step 4.2, set up forecast sample, form training sample set and test sample collection；Sample input is generally divided into following
Two classes：
A={ a1,a2,…,aK, the historical data of K renewal substitution ratio before predicting the time；
B={ b1,b2,…,bT, predict year technical factor and policy implication factor, including battery capacity, battery life,
Charging current, policy planning scaling factor；
Step 4.3, SVM forecast models are set up, following SVM regressive object functions are set up according to training sample；
Wherein xiFor the input of i-th of training sample, K (xi,xj) it is kernel function, radial direction inner product kernel function is selected herein, i.e.,Minimum majorized function is solved, is most there is excellent solution
Step 4.4, determination return decision function, willFollowing formula is substituted into, decision-making is obtained and returns letter
Number；
Step 4.5, using forecast sample and decision-making regression equation obtained in the previous step classifying type electric automobile is updated
Substitution ratio is predicted.
In a kind of above-mentioned classifying type electric automobile demand dynamic prediction method based on gray prediction and SVMs,
In the step 3, the classifying type vehicle fleet size forecast model predicted based on improved grey model and the classifying type based on SVMs
Electric automobile substitution ratio iterative model, with reference to substitution ratio model is updated, obtains classifying type electric automobile Demand Forecast Model,
Specifically：It is total using time classifying type vehicle to be predicted is obtained based on the classifying type vehicle fleet size forecast model that improved grey model is predicted
Volume data predicts the outcome, and with reference to current techniques progress and the influence of policy factor, utilizes the classifying type electricity based on SVMs
Electrical automobile substitution ratio iteration forecast model predicts that the classifying type electric automobile in corresponding time to be predicted updates substitution ratio,
Both, which predict the outcome, is combined the accurately prediction of realization dynamic：
(type of automobile)
WhereinRefer to a type electric automobiles t prediction recoverable amount,Refer to a type automobiles t prediction
Recoverable amount,Refer to a type electric automobiles t forecast updating substitution ratio.
In a kind of above-mentioned classifying type electric automobile demand dynamic prediction method based on gray prediction and SVMs,
Also include：
The step of setting up the spatial and temporal distributions forecast model of electronic private passenger car and electric taxi：Combining classification type is electronic
Automobile demand predicts the outcome, and obtains electronic private passenger car and electric taxi Demand Forecast result, based on being gone on a journey to user
Law study, realizes and the spatial and temporal distributions of electronic private passenger car and electric taxi demand is predicted, with reference to fixed traveling
The other types electric automobile of path feature, realizes the spatial and temporal distributions prediction of electric automobile demand.
In a kind of above-mentioned classifying type electric automobile demand dynamic prediction method based on gray prediction and SVMs,
It is described set up electronic private passenger car and electric taxi spatial and temporal distributions forecast model the step of space based on Trip chain it is special
Property analysis, specifically include：
Analysis condition one, space transfer：The state for defining current time is Ei, the state of subsequent time is Ej, then Ma Erke
Husband's chain can be represented with conditional probability
P(Ei→Ej)=P (Ej/Ei)=Pij
If each traveling destination is considered as into a state, according to Markov theory, the next state (purpose of vehicle
Ground) determined by current state；It is designated as pijFrom state EiSwitch to state EjState transition probability, then the transfer of one step state is general
Rate can be write as matrix form,
Wherein pijMeet following condition：
According to 4 large scenes occurred in typical Trip chain, by the way that this area's resident trip statistics is investigated, according to
The step transition probability that electric automobile drives to another destination from a destination can be just represented according to above formula is
The Origin And Destination of Trip chain is represented in matrix with english abbreviation：H represents residential block, W and represents workspace/campus
Area, SR represent shopping centre, O and represent other regions；
Analysis condition two, trip distance：The trip distance for defining traveler obeys rule distribution, according to trip distance
Destination range of choice can be reduced, particular location is determined using Monte Carlo simulation：
In formula：fDFor the probability density of trip distance；μDFor resident's average trip distance；δDFor trip distance variance,
Then spatial and temporal distributions prediction is carried out to electronic private car and electric taxi, concrete operation step includes：
Step 7.1, initialization regional cartographic information to be predicted, the input electronic private car in this area and electric taxi
Requirement forecasting estimation N；
Step 7.2, extraction origin, determine starting point；
Step 7.3, extraction destination, analyze according to trip distance, reduce range of choice, determine terminal；
Step 7.4, according to traffic information, select path；
Step 7.5, N of this area electronic private car and electric taxi be overlapped, obtain the type vehicle
Spatial and temporal distributions predict the outcome.
Therefore, the invention has the advantages that：1st, the present invention takes into full account different type Development of Electric Vehicles not
Together, electric automobile is divided into electric bus, electric taxi, electronic special-purpose vehicle, electronic officer's car and private passenger car, to dividing
Type electric automobile demand is predicted.2nd, taken into full account the factors such as government policy, technological progress to electric automobile demand
Influence, it is more accurate that electric automobile demand is predicted.3rd, the forecast model based on SVMs, can effectively be handled
Electric automobile substitutes the less situation of Belgian data, and Accurate Prediction is carried out to electric automobile demand.4th, professional etiquette is gone out with reference to user
Rule research, realizes the spatial and temporal distributions prediction of the demand of classifying type electric automobile.
Brief description of the drawings
Fig. 1 is that the classifying type electric automobile demand spatial and temporal distributions dynamic prediction with SVMs is predicted based on improved grey model
Method flow schematic diagram.
Fig. 2 is prediction classifying type vehicle fleet size schematic diagram.
Fig. 3 is the classifying type substitution ratio schematic diagram of dynamic prediction SVMs.
Fig. 4 is classifying type electric automobile demand dynamic prediction model schematic diagram.
Fig. 5 is typical Trip chain schematic diagram.
Fig. 6 is electronic private passenger car and electric taxi spatial and temporal distributions prediction schematic diagram.
Specific implementation method
First, the invention mainly includes steps：
The first step, sets up classifying type vehicle fleet size forecast model.Herein by region bus, taxi, special-purpose vehicle,
The data characteristicses analysis of the different type vehicle such as officer's car and private passenger car, selection grey method is predicted；To tradition
Grey method carries out research and is improved for its limitation and different type vehicle data feature, obtains changing for high accuracy
Enter grey forecasting model, build classifying type vehicle fleet size forecast model；By the development of the region different type vehicle obtained by investigation
Trend, by classifying type vehicle fleet size forecast model, predicts following overall development trend of the region different type vehicle.
Second step, sets up classifying type electric automobile substitution ratio iteration forecast model.Herein by region Electric Transit
The type vehicles shared by different type vehicle such as car, electric taxi, electronic special-purpose vehicle, electronic officer's car and private passenger car
Ratio data data characteristicses analysis, and influence factor non-linear behavior, selection Support vector regression method obtain decision-making
Regression equation, it is continuous using iterative method using the substitution ratio data prediction result to classifying type electric automobile of forecast sample
Amendment predicts the outcome, and builds the electric automobile classifying type substitution ratio iteration forecast model based on SVM.
3rd step, sets up classifying type electric automobile demand growth forecast model.Combining classification classifying type vehicle fleet size is predicted
Model prediction result and classifying type electric automobile substitution ratio iteration forecast model are predicted the outcome, and both are predicted the outcome according to vapour
Car type is matched, and classifying type electric automobile demand growth forecast model is set up, to classifying type electric automobile quantity in region
Realize accurate dynamic prediction.
4th step, sets up the spatial and temporal distributions forecast model of electronic private passenger car and electric taxi.Combining classification type electricity
Electrical automobile requirement forecasting result, obtains electronic private passenger car and electric taxi Demand Forecast result, based on going out to user
Row law study, realizes and the spatial and temporal distributions of electronic private passenger car and electric taxi demand is predicted, with reference to fixed row
The other types electric automobile of path feature is sailed, the spatial and temporal distributions prediction of electric automobile demand is realized.
2nd, the method for the present invention is described in detail with reference to concrete case.
1st, the classifying type vehicle fleet size forecast model based on improved grey model predicted method.
The classifying type requirement forecasting of automobile is to speculate that it is following according to past and present electric automobile statistics
Numerical value, it can be seen that the research object of the classifying type quantitative forecast of automobile is chance event, is not positive events, and vapour
Car future development uncertain factor is very more.But gray system theory can again be contained with effectively solution containing Given information
There is the forecasting problem of unknown or uncertain information system, so selection grey method is pre- to build classifying type vehicle fleet size
It is rational to survey model.
1.1st, improved grey model forecast model.
Grey Theory Forecast model be by being influenced each other between analysis system data, factor, different degree etc., utilize number
Method according to statistics, reasonable prediction is carried out to system.Its main process seeks to first do correlation analysis, and initial data is entered
Row processing, obtains randomness and reduces regular enhanced generation data, corresponding differential is set up further according to ripe generation data
Equation model, is analyzed, so that following development trend of forecasting system.Thus, it is possible to draw, as obtained by being analyzed this method
Data, can not be directly using, it is necessary to can just use after inverse generation is inverse accumulated generating.
To Traditional GM (1,1) model Limitation Analysis, can analyze place's model has two limitations：
It is very important for can be seen that processing of the first step to initial data from GM (1,1) modeling process, is
In order to which the randomness of initial data is reduced, and dispose to a certain extent " bad data ", and strengthen initial data development trend
Information.But be this mode directly by the way of Accumulating generation to the processing of data in traditional grey forecasting model
Fairly simple, " bad data " in initial data is handled to a certain extent may be not enough, and not enough " smooth ", this will for generation data
Influence the accuracys set up process, directly affect model prediction all behind grey forecasting model.
Predicting the outcome for GM (1,1) model is by setting up Grey Differential Equation to generation data, and equation solution being drawn
, then unknowm coefficient a in equation, μ accuracy will directly affect last solving result.But unknowm coefficient a, μ are asked
Solution is z with being configured with direct relation, traditional grey forecasting model directly with close to average generation sequence for background value(1)(k)=
0.5x(1)(k)+0.5x(1)(k-1) tectonic setting value, this be cause grey forecasting model predict the outcome error principal element it
One.
In summary, that is, two thinkings are improved grey forecasting model, to improve grey forecasting model prediction
Accuracy.On the one hand, it can be carried out from the angle of the processing to initial data, the method for improving generation data, increase generation number
According to smoothness, the discrete type of the improvement initial data of maximum possible, randomness, strengthen initial data development tendency, carry
The regularity of high data；On the other hand, start with from GM modeling processes, Grey Differential Equation is converted into difference side by raising
The accuracy of journey process, reduces the difference between two equations, i.e., is improved for the construction of background value.Improved grey model is predicted
Model, can be summarized as following steps：
(1), initial data smoothing processing.
According to analysis above it is recognised that to the purpose of original data processing be exactly reduce initial data discrete lines,
Randomness, the regular displaying of reinforcing initial data variation tendency, so being highly desirable to enter the processing procedure of initial data
Row is improved so that generation data light slippery is more preferable.Then generation formula is improved as follows：
Original data sequence two ends data smoothing processing mode：
x(1)(1)=(3x(0)(1)+x(0)(2))/4
x(1)(n)=(x(0)(n-1)+3x(0)(n))/4
Original data sequence intermediate data smoothing processing mode：
x(1)(m)=(x(0)(m-1)+2x(0)(m)+x(0)(m+1))/41 ＜ m ＜ n
, instead of traditional Accumulating generation mode, more smooth data can be obtained using above processing method.
(2) Grey Differential Equation, is set up.
Model parameter vector α is obtained with least square method i.e.Obtain GM (1,1) model；
(3), using rear backdrop value is improved, difference equation is set up.
The selection of background value, is directly connected to the accuracy that the differential equation is converted into difference equation, for directly with tight
Adjacent average generation sequence replaces background value to be inaccurate, so the amendment for background value is very important.Substantially
For, z(1)(k) it is x(1)(t) approximate integral value, i.e.,So directly with close to average generation sequence
It is to be grossly inaccurate to be used as background value.In order to be able to which the differential equation is accurately converted into difference equation, for high growth, put down
Increase, the difference equation formula of low growth should have differences.Obviously make for flat growth, the difference equation of low growth
With smaller as background value error close to average generation sequence, but for high growth difference equation there is with this method compared with
Big the problem of, so being modified here mainly for the background value of the difference equation of high growth.
For cumulative data exponential form, i.e. x can be processed data into the situation of high growth trend(1)(t)=μ eat,
The formula is brought into integral formula, i.e.,It can obtain
For it is flat increase or it is incremental it is long using close to formation sequence as background value, i.e.,：
z(1)(k)=0.5x(1)(k)+0.5x(1)(k-1)
(4), model is tested.
The model having built up is tested using test set, judgment models are suitable, if suitable feasible, are entered
Row next step；If being unsatisfactory for error requirements, the first step is returned to, data are carried out with a smoothing processing again；
(5), inverse growth is predicted the outcome.
The result obtained is reduced, that is, carries out inverse growth process (i.e. inverse accumulated generating IAGO), herein will be according to cumulative
The number of times of generation is carried out, and is really predicted the outcome.
1.2nd, classifying type vehicle fleet size forecast model.
Using the classifying type vehicle historical data in former years as input quantity, the pre- quantitation of classifying type vehicle is output quantity, is based on
Improved grey model forecast model GM (1,1), builds classifying type vehicle fleet size forecast model.Implementation steps are summarized as follows：
The first step, is handled initial data：It regard the historical data of the vehicle of a certain type in recent years as original number
According to using the Accumulating generation method after improvement, being smoothed to initial data.
Second step, selects suitable background value, the differential equation is accurately converted into difference equation：According to the first step to original
The processing of data, is analyzed data substantially growth trend, if initial data increases or low growth to be flat, using original
Method using background value is replaced close to formation sequence, is further calculated；If initial data is high growth trend, improvement is utilized
Background value generation formula, further calculating is done after bringing into.
3rd step, using least square method, obtains parameter vector α to be asked.
4th step, tests to the model having built up, and judges whether the solution of the differential equation is suitable, if properly, entering
Row next step, otherwise returns to the first step, to the further processing of data.
5th step, is predicted to result according to the suitable difference equation of foundation, i.e., difference equation is solved.
6th step, is reduced to the result obtained, that is, carries out inverse growth process (i.e. inverse accumulated generating), herein will be according to tired
Plus the number of times of generation is carried out, predicting the outcome for certain final type of vehicle quantity is obtained.
It is as shown in Figure 2 that classifying type vehicle fleet size forecast model sets up schematic diagram.
2nd, the classifying type electric automobile substitution ratio dynamic prediction model based on SVMs.
SVMs is a kind of very new method, but the ability of this method process problem is but defeated by not at all
Traditional Forecasting Methodology, is a kind of very useful method.SVMs be based on traditional Statistical Learning Theory, from
Optimal face development of categories in the case of linear separability and come.The core of this method is the method using machine learning, finds
The fitting function of each point, is accurately predicted in plane.
SVMs is for solving the problems such as finite sample, non-linear and high dimensional pattern are recognized with very big advantage.
Can be in the case where data loudness be not enough using SVMs, it is possible to use small sample preferably searches out rule therein
Rule, accurately to be predicted.There are the feelings such as the smaller, data deficiencies of sample because classifying type electric automobile updates ratio data
Condition, selects the method for SVMs come estimation is predicted to it is rational.
2.1st, SVMs iteration forecast model.
The classifying type electric automobile substitution ratio iteration forecast model of SVMs is established herein, for Classical forecast
The defect of method, this paper presents include the influence factor such as technological progress and government policy guiding.Input vector x ∈ R in modelm
For electric automobile substitution ratio historical data, technological progress influence factor, government policy guiding etc., output y substitutes to update
Scale prediction value.Training sample set and forecast sample collection are set up according to given data, SVM regressive object functions is set up, solves most
Excellent solution brings back to recurrence decision function equation, obtains regression forecasting function, finally calculates and predicts the outcome.
(1) historical data is pre-processed and normalized.
WhereinFor the data value after normalization, xiFor actual, historical data, xi min=min (xi), xi max=max (xi),
M is the dimension of input vector, that is, influences the number of predictive factorses.
(2) forecast sample is set up, training sample set and test sample collection is formed.Sample input is generally divided into two categories below：
A={ a1,a2,…,aK, the historical data of K renewal substitution ratio before predicting the time；
B={ b1,b2,…,bT, predict year technical factor and policy implication factor, including battery capacity, battery life,
The factors such as charging current, policy planning ratio (data after normalization)；
(3) SVM forecast models are set up, following SVM regressive object functions are set up according to training sample.
Wherein xiFor the input of i-th of training sample, K (xi,xj) it is kernel function, radial direction inner product kernel function is selected herein, i.e.,Minimum majorized function is solved, is most there is excellent solution
(4) determine to return decision function, willFollowing formula is substituted into, decision-making regression function is obtained.
(5) classifying type electric automobile is updated using forecast sample and decision-making regression equation obtained in the previous step and substitutes ratio
Example is predicted.
Classifying type electric automobile substitution ratio iterative model concrete operations flow as shown in Figure 3 based on SVMs
2.2nd, SVMs iteration forecast model prediction classifying type electric automobile updates the advantage of substitution ratio.
SVM has a clear superiority in solution finite sample, non-linear and higher-dimension identification problem, and influences to update substitution ratio
Factor there is non-linear behavior, therefore the Nonlinear Learning ability and estimated performance superior using SVMs herein is proposed
Classifying type electric automobile substitution ratio iteration forecast model based on SVMs.
(1) SVM uses structural risk minimization principle, and the study of crossing for successfully solving traditional neural network is asked with deficient study
Topic, therefore SVM is applied to electric automobile and updated in substitution ratio prediction, conventional method can be avoided to use neural network prediction institute
That brings crosses study phenomenon.
(2) SVM is returned by kernel function, and the low-dimensional input space is mapped into high-dimensional feature space, then empty in high dimensional feature
Between in carry out linear regression.Algorithm is finally converted into as a quadratic form optimization problem, and in theory, what is obtained is complete
Office's optimal solution.Compared with the neutral net used in traditional prediction method, unavoidable local extremum problem is solved.
(3) SVM uses structural risk minimization principle, improves the general ability to following sample.Therefore use be based on to
The iteration forecast model of amount machine effectively can be predicted to electric automobile classifying type substitution ratio.
3rd, classifying type electric automobile demand dynamic prediction model.
This part is with reference to the classifying type vehicle fleet size forecast model predicted based on improved grey model and based on supporting vector herein
The classifying type electric automobile substitution ratio iterative model of machine, with reference to substitution ratio model is updated, obtaining classifying type electric automobile needs
Seek forecast model.
Time classifying type to be predicted is obtained first with based on the classifying type vehicle fleet size forecast model that improved grey model is predicted
Vehicle conceptual data predicts the outcome, and with reference to current techniques progress and the influence of policy factor, utilizes point based on SVMs
Type electric automobile substitution ratio iteration forecast model is predicted that the classifying type electric automobile in corresponding time to be predicted updates and replaced
For ratio, both, which predict the outcome, is combined the accurately prediction of realization dynamic：
(type of automobile)
WhereinRefer to a type electric automobiles t prediction recoverable amount,Refer to a type automobiles t prediction
Recoverable amount,Refer to a type electric automobiles t forecast updating substitution ratio.
Concrete operations logic as shown in Figure 4
4th, electric automobile spatial and temporal distributions are predicted.
Electric bus, electronic special-purpose vehicle have the space-time that the driving path and running time more fixed more are fixed
Distribution, the spatial and temporal distributions of electronic private passenger car and electric taxi and the driving habit of automobile user have close connection
System, however electric automobile spatial and temporal distributions prediction to the requirement forecasting of electric automobile charging and conversion electric, electric automobile traffic path recommend and
The researchs such as electric automobile electric charging station planning are all significant, so being highly desirable to electronic private passenger car and electronic
The spatial and temporal distributions of taxi are studied.
First, to traveler behavioral characteristic, driving habit is studied, and analyzes the travel destination and starting point class of traveler
The potential relation of type.Secondly, the specific beginning and end position that traveler is gone on a journey is determined with reference to actual traffic information, it is determined that trip
Route, most electronic private passenger car and electric taxi all in the region is overlapped at last, obtains the region electronic
The electronic private car of automobile space-time and the spatial and temporal distributions situation of electric taxi.
4.1st, trip link analysis.
Trip chain refers to from initially, by going on a journey several times, finally turns again to whole process initially.This
The traveler travel activity of one day can be handled by planting the method for trip chain type, with daily life rule basic one
Cause, it is very close with traveler practice decision process, it can effectively simulate traveler and each is gone on a journey during Trip chain
The decision process of activity, so as to be provided a great help for the trip rule for analyzing automobile user.Can be with by investigation
Draw, typical Trip chain is as shown in Figure 5：
4.2nd, the spatial character analysis of Trip chain.
(1) space transfer.
Markoff process is for describing the random process with markov property：If per next state transfer only with it is previous
The state at quarter is unrelated about and with past state, and discrete markoff process is referred to as Markov Chain.Remember the shape at current time
State is Ei, the state of subsequent time is Ej, then Markov Chain can be represented with conditional probability
P(Ei→Ej)=P (Ej/Ei)=Pij
If each traveling destination is considered as into a state, according to Markov theory, the next state (purpose of vehicle
Ground) determined by current state.It is designated as pijFrom state EiSwitch to state EjState transition probability, then the transfer of one step state is general
Rate can be write as matrix form,
Wherein pijMeet following condition：
According to 4 large scenes occurred in typical Trip chain, by the way that this area's resident trip statistics is investigated, according to
The step transition probability that electric automobile drives to another destination from a destination can be just represented according to above formula is
The Origin And Destination of Trip chain is represented in matrix with english abbreviation：H represents residential block, W and represents workspace/campus
Area, SR represent shopping centre, O and represent other regions.
(2) trip distance.
Show that the trip distance of traveler is obeyed rule and is distributed by research, can be by destination according to trip distance
Range of choice is reduced, and particular location is determined using Monte Carlo simulation.
In formula：fDFor the probability density of trip distance；μDFor resident's average trip distance；δDFor trip distance variance.
4.3rd, electronic private car and the prediction of electric taxi spatial and temporal distributions.
According to researching and analysing above, spatial and temporal distributions prediction, tool can be carried out to electronic private car and electric taxi
Body operating procedure is as shown in Figure 6：
The first step, initializes regional cartographic information to be predicted, the input electronic private car in this area and electric taxi
Requirement forecasting estimation N；
Second step, extracts origin, determines starting point；
3rd step, extracts destination, is analyzed according to trip distance, reduces range of choice, determines terminal；
4th step, according to traffic information, selects path；
5th step, N of this area electronic private car and electric taxi are overlapped, the type vehicle is obtained
Spatial and temporal distributions predict the outcome.
4.4th, electric automobile spatial and temporal distributions are predicted.
By analysis above it is known that the running time and driving path right and wrong of electric bus and electronic special-purpose vehicle
Often fixed, i.e., the vehicle of this two type has the spatial and temporal distributions characteristic more fixed, it is possible to by by this two classes car
Distribution character predicts the outcome with the spatial and temporal distributions of electronic private passenger car and electric taxi be just superimposed, to obtain electronic vapour
The spatial and temporal distributions prediction of car.
It should be appreciated that the part that this specification is not elaborated belongs to prior art.It should be appreciated that above-mentioned
Description for preferred embodiment is more detailed, therefore can not be considered the limitation to scope of patent protection of the present invention, this
The those of ordinary skill in field is under the enlightenment of the present invention, in the case where not departing from the ambit that the claims in the present invention are protected,
Replacement can also be made or deformed, each fallen within protection scope of the present invention, the scope that is claimed of the invention should be with appended
Claim is defined.
Claims (7)
1. a kind of classifying type electric automobile demand dynamic prediction method based on gray prediction and SVMs, its feature exists
In, including：
Step 1, set up classifying type vehicle fleet size forecast model, by region bus, taxi, special-purpose vehicle, officer's car and
The data characteristicses analysis of the different type vehicle such as private passenger car, selection grey method is predicted；To traditional gray prediction
Method carries out research and is improved for its limitation and different type vehicle data feature, and the improved grey model for obtaining high accuracy is pre-
Model is surveyed, classifying type vehicle fleet size forecast model is built；By the development trend of the region different type vehicle obtained by investigation, pass through
Classifying type vehicle fleet size forecast model, predicts following overall development trend of the region different type vehicle；
Step 2, set up classifying type electric automobile substitution ratio iteration forecast model, by region electric bus, it is electronic go out
Hire a car, the ratio data of the type vehicle shared by different type vehicle such as electronic special-purpose vehicle, electronic officer's car and private passenger car
Data characteristicses analysis, and influence factor non-linear behavior, selection Support vector regression method obtain decision-making regression equation,
Using the substitution ratio data prediction result to classifying type electric automobile of forecast sample, prediction knot is constantly corrected using iterative method
Really, the electric automobile classifying type substitution ratio iteration forecast model based on SVM is built；
Step 3, classifying type electric automobile demand growth forecast model is set up；Combining classification classifying type vehicle fleet size forecast model is pre-
Survey result and classifying type electric automobile substitution ratio iteration forecast model predicts the outcome, both are predicted the outcome according to car category
Matched, set up classifying type electric automobile demand growth forecast model, classifying type electric automobile quantity in region is realized accurate
True dynamic prediction.
2. a kind of classifying type electric automobile demand dynamic based on gray prediction and SVMs according to claim 1
Forecasting Methodology, it is characterised in that in the step 1, sets up improved grey model forecast model, including：
Step 2.1, initial data smoothing step, are improved generation formula as follows：
Original data sequence two ends data smoothing processing mode：
x(1)(1)=(3x(0)(1)+x(0)(2))/4
x(1)(n)=(x(0)(n-1)+3x(0)(n))/4
Original data sequence intermediate data smoothing processing mode：
x(1)(m)=(x(0)(m-1)+2x(0)(m)+x(0)(m+1))/4 1 ＜ m ＜ n
, instead of traditional Accumulating generation mode, more smooth data can be obtained using above processing method；
Step 2.2, Grey Differential Equation is set up, based on equation below：
<mrow>
<mfrac>
<mrow>
<msup>
<mi>dX</mi>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</msup>
</mrow>
<mrow>
<mi>d</mi>
<mi>t</mi>
</mrow>
</mfrac>
<mo>+</mo>
<msup>
<mi>aX</mi>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</msup>
<mo>=</mo>
<mi>&mu;</mi>
</mrow>
Model parameter vector α is obtained with least square method i.e.Obtain GM (1,1) model；
Step 2.3, using improving rear backdrop value, the step of setting up difference equation, for the background value of the difference equation of high growth
It is modified；
For cumulative data exponential form, i.e. x can be processed data into the situation of high growth trend(1)(t)=μ eat, by this
Formula is brought into integral formula, i.e.,It can obtain
<mrow>
<msup>
<mi>z</mi>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</msup>
<mrow>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<munderover>
<mo>&Integral;</mo>
<mrow>
<mi>k</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
<mi>k</mi>
</munderover>
<msup>
<mi>x</mi>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</msup>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mi>d</mi>
<mi>t</mi>
<mo>=</mo>
<munderover>
<mo>&Integral;</mo>
<mrow>
<mi>k</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
<mi>k</mi>
</munderover>
<msup>
<mi>&mu;e</mi>
<mrow>
<mi>a</mi>
<mi>t</mi>
</mrow>
</msup>
<mi>d</mi>
<mi>t</mi>
<mo>=</mo>
<mfrac>
<mi>&mu;</mi>
<mi>a</mi>
</mfrac>
<mrow>
<mo>(</mo>
<mrow>
<msup>
<mi>e</mi>
<mrow>
<mi>a</mi>
<mi>k</mi>
</mrow>
</msup>
<mo>-</mo>
<msup>
<mi>e</mi>
<mrow>
<mi>a</mi>
<mrow>
<mo>(</mo>
<mrow>
<mi>k</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
<mo>)</mo>
</mrow>
</mrow>
</msup>
</mrow>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mrow>
<msup>
<mi>x</mi>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</msup>
<mrow>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
</mrow>
<mo>-</mo>
<msup>
<mi>x</mi>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</msup>
<mrow>
<mo>(</mo>
<mrow>
<mi>k</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mi>ln</mi>
<mi> </mi>
<msup>
<mi>x</mi>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</msup>
<mrow>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mi>ln</mi>
<mi> </mi>
<msup>
<mi>x</mi>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</msup>
<mrow>
<mo>(</mo>
<mrow>
<mi>k</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
<mo>)</mo>
</mrow>
</mrow>
</mfrac>
</mrow>
For it is flat increase or it is incremental it is long using close to formation sequence as background value, i.e.,：
z(1)(k)=0.5x(1)(k)+0.5x(1)(k-1)
Step 2.4, the step of tested to model：The model having built up is tested, test set testing model is utilized
It is whether suitable, if suitable feasible, carry out next step；If being unsatisfactory for error requirements, the first step is returned to, data are entered again
Otherwise smoothing processing of row returns to the first step, and data are carried out with a smoothing processing again using formula；
Step 2.5, inverse growth are predicted the outcome：The result obtained is reduced, inverse growth process is carried out, that is, it is i.e. tired
Subtract generation IAGO, to be carried out, really be predicted the outcome according to the number of times of Accumulating generation herein.
3. a kind of classifying type electric automobile demand dynamic based on gray prediction and SVMs according to claim 1
Forecasting Methodology, it is characterised in that in the step 1, the foundation of classifying type vehicle fleet size forecast model includes：
Step 3.1, initial data is handled：Using the historical data of the vehicle of a certain type in recent years as initial data,
Using the Accumulating generation method after improvement, initial data is smoothed；
Step 3.2, the suitable background value of selection, difference equation is accurately converted into by the differential equation：According to the first step to original number
According to processing, data substantially growth trend is analyzed, if initial data increases or low growth to be flat, original-party is utilized
Method using background value is replaced close to formation sequence, is further calculated；If initial data is high growth trend, using improved
Background value generates formula, and further calculating is done after bringing into；
Step 3.3, using least square method, obtain parameter vector α to be asked；
Step 3.4, the model having built up is tested, judge whether the solution of the differential equation is suitable, if properly, under carrying out
One step, otherwise returns to the first step, to the further processing of data；
Step 3.5, according to the suitable difference equation of foundation result is predicted, i.e., difference equation solved；
Step 3.6, the result obtained is reduced, inverse growth process, that is, inverse accumulated generating are carried out, according to Accumulating generation
Number of times is carried out, and obtains predicting the outcome for certain final type of vehicle quantity.
4. a kind of classifying type electric automobile demand dynamic based on gray prediction and SVMs according to claim 1
In Forecasting Methodology, it is characterised in that in the step 2, the SVMs iteration forecast model, input vector x ∈ RmFor
Electric automobile substitution ratio historical data, technological progress influence factor, government policy guiding etc., output y substitutes ratio to update
Example predicted value；Training sample set and forecast sample collection are set up according to given data, SVM regressive object functions is set up, solves optimal
Solution brings back to recurrence decision function equation, obtains regression forecasting function, finally calculates and predicts the outcome, specifically includes：
Step 4.1, historical data is pre-processed and normalized；
<mrow>
<mover>
<mi>x</mi>
<mo>&CenterDot;</mo>
</mover>
<mo>=</mo>
<mfrac>
<mrow>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
<mo>-</mo>
<msub>
<mi>x</mi>
<mrow>
<mi>i</mi>
<mi>min</mi>
</mrow>
</msub>
</mrow>
<mrow>
<msub>
<mi>x</mi>
<mrow>
<mi>i</mi>
<mi>max</mi>
</mrow>
</msub>
<mo>-</mo>
<msub>
<mi>x</mi>
<mrow>
<mi>i</mi>
<mi>min</mi>
</mrow>
</msub>
</mrow>
</mfrac>
<mo>,</mo>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
<mo>,</mo>
<mn>2</mn>
<mo>,</mo>
<mo>...</mo>
<mo>,</mo>
<mi>m</mi>
<mo>)</mo>
</mrow>
</mrow>
WhereinFor the data value after normalization, xiFor actual, historical data, ximin=min (xi), ximax=max (xi), m is defeated
The dimension of incoming vector, that is, influence the number of predictive factorses；
Step 4.2, set up forecast sample, form training sample set and test sample collection；Sample input is generally divided into two categories below：
A={ a1,a2,…,aK, the historical data of K renewal substitution ratio before predicting the time；
B={ b1,b2,…,bT, predict the technical factor and policy implication factor in year, including battery capacity, battery life, charging
Electric current, policy planning scaling factor；
Step 4.3, SVM forecast models are set up, following SVM regressive object functions are set up according to training sample；
<mrow>
<mi>min</mi>
<mi> </mi>
<mi>i</mi>
<mi>m</mi>
<mi>i</mi>
<mi>z</mi>
<mi>e</mi>
<mrow>
<mo>{</mo>
<mrow>
<mo>-</mo>
<munderover>
<mi>&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>n</mi>
</munderover>
<msub>
<mi>y</mi>
<mi>i</mi>
</msub>
<mrow>
<mo>(</mo>
<mrow>
<msubsup>
<mi>a</mi>
<mi>j</mi>
<mo>*</mo>
</msubsup>
<mo>-</mo>
<msub>
<mi>a</mi>
<mi>i</mi>
</msub>
</mrow>
<mo>)</mo>
</mrow>
<mo>+</mo>
<mi>&epsiv;</mi>
<munderover>
<mi>&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>n</mi>
</munderover>
<mrow>
<mo>(</mo>
<mrow>
<msub>
<mi>a</mi>
<mi>i</mi>
</msub>
<mo>+</mo>
<msubsup>
<mi>a</mi>
<mi>i</mi>
<mo>*</mo>
</msubsup>
</mrow>
<mo>)</mo>
</mrow>
<mo>+</mo>
<mfrac>
<mn>1</mn>
<mn>2</mn>
</mfrac>
<munderover>
<mi>&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>n</mi>
</munderover>
<munderover>
<mi>&Sigma;</mi>
<mrow>
<mi>j</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>n</mi>
</munderover>
<mrow>
<mo>(</mo>
<mrow>
<msubsup>
<mi>a</mi>
<mi>i</mi>
<mo>*</mo>
</msubsup>
<mo>-</mo>
<msub>
<mi>a</mi>
<mi>i</mi>
</msub>
</mrow>
<mo>)</mo>
</mrow>
<mrow>
<mo>(</mo>
<mrow>
<msubsup>
<mi>a</mi>
<mi>j</mi>
<mo>*</mo>
</msubsup>
<mo>-</mo>
<msub>
<mi>a</mi>
<mi>j</mi>
</msub>
</mrow>
<mo>)</mo>
</mrow>
<mi>K</mi>
<mrow>
<mo>(</mo>
<mrow>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
<mo>,</mo>
<msub>
<mi>x</mi>
<mi>j</mi>
</msub>
</mrow>
<mo>)</mo>
</mrow>
</mrow>
<mo>}</mo>
</mrow>
</mrow>
Wherein xiFor the input of i-th of training sample, K (xi,xj) it is kernel function, radial direction inner product kernel function is selected herein, i.e.,Minimum majorized function is solved, is most there is excellent solution
Step 4.4, determination return decision function, willFollowing formula is substituted into, decision-making regression function is obtained；
<mrow>
<mi>f</mi>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mi>w</mi>
<mo>&CenterDot;</mo>
<mi>x</mi>
<mo>+</mo>
<mi>b</mi>
<mo>=</mo>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>k</mi>
</munderover>
<mrow>
<mo>(</mo>
<msub>
<mi>a</mi>
<mi>i</mi>
</msub>
<mo>-</mo>
<msubsup>
<mi>a</mi>
<mi>i</mi>
<mo>*</mo>
</msubsup>
<mo>)</mo>
</mrow>
<mi>K</mi>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>,</mo>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>+</mo>
<mi>b</mi>
</mrow>
Step 4.5, classifying type electric automobile is updated using forecast sample and decision-making regression equation obtained in the previous step substituted
Ratio is predicted.
5. a kind of classifying type electric automobile demand dynamic based on gray prediction and SVMs according to claim 1
Forecasting Methodology, it is characterised in that in the step 3, the classifying type vehicle fleet size forecast model and base predicted based on improved grey model
In the classifying type electric automobile substitution ratio iterative model of SVMs, with reference to substitution ratio model is updated, classifying type is obtained
Electric automobile Demand Forecast Model, be specifically：Obtained using based on the classifying type vehicle fleet size forecast model that improved grey model is predicted
Time classifying type vehicle conceptual data to be predicted predicts the outcome, and with reference to current techniques progress and the influence of policy factor, utilizes base
The classification in corresponding time to be predicted is predicted in the classifying type electric automobile substitution ratio iteration forecast model of SVMs
Type electric automobile updates substitution ratio, and both, which predict the outcome, is combined the accurately prediction of realization dynamic：
(type of automobile)
WhereinRefer to a type electric automobiles t prediction recoverable amount,Refer to that a type automobiles t prediction is possessed
Amount,Refer to a type electric automobiles t forecast updating substitution ratio.
6. a kind of classifying type electric automobile demand dynamic based on gray prediction and SVMs according to claim 1
Forecasting Methodology, it is characterised in that also include：
The step of setting up the spatial and temporal distributions forecast model of electronic private passenger car and electric taxi：Combining classification type electric automobile
Requirement forecasting result, obtains electronic private passenger car and electric taxi Demand Forecast result, based on rule of being gone on a journey to user
Research, realizes and the spatial and temporal distributions of electronic private passenger car and electric taxi demand is predicted, with reference to fixed driving path
The other types electric automobile of feature, realizes the spatial and temporal distributions prediction of electric automobile demand.
7. a kind of classifying type electric automobile demand dynamic based on gray prediction and SVMs according to claim 6
Forecasting Methodology, it is characterised in that the step of the spatial and temporal distributions forecast model for setting up electronic private passenger car and electric taxi
Suddenly the spatial character analysis based on Trip chain, is specifically included：
Analysis condition one, space transfer：The state for defining current time is Ei, the state of subsequent time is Ej, then Markov Chain
It can be represented with conditional probability
P(Ei→Ej)=P (Ej/Ei)=Pij
If each traveling destination is considered as into a state, according to Markov theory, the next state of vehicle (destination) is
Determined by current state；It is designated as pijFrom state EiSwitch to state EjState transition probability, then one step state transition probability can
Being write as matrix form is,
Wherein pijMeet following condition：
<mfenced open = "{" close = "">
<mtable>
<mtr>
<mtd>
<mrow>
<mn>0</mn>
<mo><</mo>
<msub>
<mi>p</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mo><</mo>
<mn>1</mn>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
<mo>=</mo>
<mn>1</mn>
<mo>,</mo>
<mn>2</mn>
<mo>...</mo>
<mi>n</mi>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>j</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>n</mi>
</munderover>
<msub>
<mi>p</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>=</mo>
<mn>1</mn>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
<mo>,</mo>
<mn>2</mn>
<mo>...</mo>
<mi>n</mi>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
According to 4 large scenes occurred in typical Trip chain, by the way that this area's resident trip statistics is investigated, according to upper
Formula can just represent the step transition probability that electric automobile drives to another destination from a destination
The Origin And Destination of Trip chain is represented in matrix with english abbreviation：H represent residential block, W represent workspace/campus area,
SR represents shopping centre, O and represents other regions；
Analysis condition two, trip distance：The trip distance for defining traveler obeys rule distribution, can be with according to trip distance
Destination range of choice is reduced, particular location is determined using Monte Carlo simulation：
<mrow>
<msub>
<mi>f</mi>
<mi>D</mi>
</msub>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mrow>
<msub>
<mi>x&delta;</mi>
<mi>D</mi>
</msub>
<msqrt>
<mrow>
<mn>2</mn>
<mi>&pi;</mi>
</mrow>
</msqrt>
</mrow>
</mfrac>
<mi>exp</mi>
<mrow>
<mo>(</mo>
<mo>-</mo>
<mfrac>
<msup>
<mrow>
<mo>(</mo>
<mi>ln</mi>
<mi> </mi>
<mi>x</mi>
<mo>-</mo>
<msub>
<mi>&mu;</mi>
<mi>D</mi>
</msub>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mrow>
<mn>2</mn>
<msubsup>
<mi>&delta;</mi>
<mi>D</mi>
<mn>2</mn>
</msubsup>
</mrow>
</mfrac>
<mo>)</mo>
</mrow>
</mrow>
In formula：fDFor the probability density of trip distance；μDFor resident's average trip distance；δDFor trip distance variance,
Then spatial and temporal distributions prediction is carried out to electronic private car and electric taxi, concrete operation step includes：
Step 7.1, the need for initializing regional cartographic information to be predicted, the input electronic private car in this area and electric taxi
Seek predicted estimate N；
Step 7.2, extraction origin, determine starting point；
Step 7.3, extraction destination, analyze according to trip distance, reduce range of choice, determine terminal；
Step 7.4, according to traffic information, select path；
Step 7.5, N of this area electronic private car and electric taxi be overlapped, obtain the type vehicle when
Space division cloth predicts the outcome.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710297707.7A CN107146013A (en) | 2017-04-28 | 2017-04-28 | A kind of classifying type electric automobile demand spatial and temporal distributions dynamic prediction method based on gray prediction and SVMs |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710297707.7A CN107146013A (en) | 2017-04-28 | 2017-04-28 | A kind of classifying type electric automobile demand spatial and temporal distributions dynamic prediction method based on gray prediction and SVMs |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107146013A true CN107146013A (en) | 2017-09-08 |
Family
ID=59774467
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710297707.7A Pending CN107146013A (en) | 2017-04-28 | 2017-04-28 | A kind of classifying type electric automobile demand spatial and temporal distributions dynamic prediction method based on gray prediction and SVMs |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107146013A (en) |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108510128A (en) * | 2018-04-11 | 2018-09-07 | 华南理工大学广州学院 | A kind of region electric vehicle charging load spatial and temporal distributions prediction technique |
| CN108681795A (en) * | 2018-05-23 | 2018-10-19 | 华南理工大学 | Electric vehicle charging load spatio-temporal prediction method under urban road network and the constraint of user's Trip chain |
| CN109102114A (en) * | 2018-08-02 | 2018-12-28 | 中南大学 | A kind of bus trip get-off stop estimation method based on data fusion |
| CN109308541A (en) * | 2018-08-02 | 2019-02-05 | 华南理工大学 | A kind of electric car charging load forecasting method based on scene analysis |
| CN109711913A (en) * | 2017-10-25 | 2019-05-03 | 中国电力科学研究院有限公司 | A kind of electric car supply and demand difference prediction technique and charging pile method of allocation plan |
| WO2019105065A1 (en) * | 2017-11-28 | 2019-06-06 | 蔚来汽车有限公司 | Electronic map based charging request initiating time prediction method and device |
| CN110046787A (en) * | 2019-01-15 | 2019-07-23 | 重庆邮电大学 | A kind of urban area charging demand for electric vehicles spatio-temporal prediction method |
| CN110287543A (en) * | 2019-05-30 | 2019-09-27 | 武汉大学 | Method for predicting service life of relay protection device |
| CN110348644A (en) * | 2019-07-18 | 2019-10-18 | 广州市交通规划研究院 | A kind of method of electric car electrically-charging equipment construction demand prediction |
| CN111310990A (en) * | 2020-02-07 | 2020-06-19 | 北京联合大学 | Improved gray combination model-based track quality prediction method and system |
| CN112348258A (en) * | 2020-11-09 | 2021-02-09 | 合肥工业大学 | Shared bicycle predictive scheduling method based on deep Q network |
| CN112561499A (en) * | 2021-02-01 | 2021-03-26 | 开封大学 | New energy data management system based on low-carbon economy |
| CN113077007A (en) * | 2021-04-19 | 2021-07-06 | 国网浙江省电力有限公司物资分公司 | Model matching system for power material configuration demand measurement |
| CN113379159A (en) * | 2021-07-05 | 2021-09-10 | 南京工程学院 | Taxi driver passenger-searching route recommendation method based on gray model and Markov decision process |
-
2017
- 2017-04-28 CN CN201710297707.7A patent/CN107146013A/en active Pending
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109711913A (en) * | 2017-10-25 | 2019-05-03 | 中国电力科学研究院有限公司 | A kind of electric car supply and demand difference prediction technique and charging pile method of allocation plan |
| WO2019105065A1 (en) * | 2017-11-28 | 2019-06-06 | 蔚来汽车有限公司 | Electronic map based charging request initiating time prediction method and device |
| CN108510128A (en) * | 2018-04-11 | 2018-09-07 | 华南理工大学广州学院 | A kind of region electric vehicle charging load spatial and temporal distributions prediction technique |
| CN108681795B (en) * | 2018-05-23 | 2021-12-21 | 华南理工大学 | Electric automobile charging load space-time prediction method under constraint of urban traffic network and user travel chain |
| CN108681795A (en) * | 2018-05-23 | 2018-10-19 | 华南理工大学 | Electric vehicle charging load spatio-temporal prediction method under urban road network and the constraint of user's Trip chain |
| CN109102114A (en) * | 2018-08-02 | 2018-12-28 | 中南大学 | A kind of bus trip get-off stop estimation method based on data fusion |
| CN109308541A (en) * | 2018-08-02 | 2019-02-05 | 华南理工大学 | A kind of electric car charging load forecasting method based on scene analysis |
| CN109102114B (en) * | 2018-08-02 | 2022-04-05 | 中南大学 | Bus trip getting-off station estimation method based on data fusion |
| CN110046787A (en) * | 2019-01-15 | 2019-07-23 | 重庆邮电大学 | A kind of urban area charging demand for electric vehicles spatio-temporal prediction method |
| CN110287543A (en) * | 2019-05-30 | 2019-09-27 | 武汉大学 | Method for predicting service life of relay protection device |
| CN110287543B (en) * | 2019-05-30 | 2022-05-13 | 武汉大学 | Method for predicting service life of relay protection device |
| CN110348644A (en) * | 2019-07-18 | 2019-10-18 | 广州市交通规划研究院 | A kind of method of electric car electrically-charging equipment construction demand prediction |
| CN110348644B (en) * | 2019-07-18 | 2021-03-30 | 广州市交通规划研究院 | Method for predicting construction requirements of electric vehicle charging facilities |
| CN111310990A (en) * | 2020-02-07 | 2020-06-19 | 北京联合大学 | Improved gray combination model-based track quality prediction method and system |
| CN112348258A (en) * | 2020-11-09 | 2021-02-09 | 合肥工业大学 | Shared bicycle predictive scheduling method based on deep Q network |
| CN112348258B (en) * | 2020-11-09 | 2022-09-20 | 合肥工业大学 | Shared bicycle predictive scheduling method based on deep Q network |
| CN112561499A (en) * | 2021-02-01 | 2021-03-26 | 开封大学 | New energy data management system based on low-carbon economy |
| CN113077007A (en) * | 2021-04-19 | 2021-07-06 | 国网浙江省电力有限公司物资分公司 | Model matching system for power material configuration demand measurement |
| CN113379159A (en) * | 2021-07-05 | 2021-09-10 | 南京工程学院 | Taxi driver passenger-searching route recommendation method based on gray model and Markov decision process |
| CN113379159B (en) * | 2021-07-05 | 2024-01-02 | 南京工程学院 | Taxi driver passenger searching route recommendation method based on gray model and Markov decision process |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107146013A (en) | A kind of classifying type electric automobile demand spatial and temporal distributions dynamic prediction method based on gray prediction and SVMs | |
| CN108761509B (en) | Automobile driving track and mileage prediction method based on historical data | |
| Sun et al. | A machine learning method for predicting driving range of battery electric vehicles | |
| Lin et al. | An ensemble learning velocity prediction-based energy management strategy for a plug-in hybrid electric vehicle considering driving pattern adaptive reference SOC | |
| CN103745110B (en) | Method of estimating operational driving range of all-electric buses | |
| Shankar et al. | Method for estimating the energy consumption of electric vehicles and plug‐in hybrid electric vehicles under real‐world driving conditions | |
| Li et al. | Prediction of electric bus energy consumption with stochastic speed profile generation modelling and data driven method based on real-world big data | |
| Xie et al. | A hybrid method combining Markov prediction and fuzzy classification for driving condition recognition | |
| CN106427589A (en) | Electric car driving range estimation method based on prediction of working condition and fuzzy energy consumption | |
| CN113902011A (en) | Urban rail transit short-time passenger flow prediction method based on cyclic neural network | |
| CN103745111A (en) | Method of predicting driving range of all-electric passenger vehicles | |
| CN106910199A (en) | Towards the car networking mass-rent method of city space information gathering | |
| CN111523714B (en) | Site selection layout method and device for electric power charging station | |
| CN109934403A (en) | Charge load Analysis prediction technique in electric car resident region based on mathematical model | |
| Guo et al. | A novel energy consumption prediction model with combination of road information and driving style of BEVs | |
| CN112734094B (en) | Intelligent city intelligent rail vehicle fault gene prediction method and system | |
| CN114492919A (en) | Electric automobile charging load prediction system and method based on traffic flow | |
| CN114048920A (en) | Site selection layout method, device, equipment and storage medium for charging facility construction | |
| CN112884014A (en) | Traffic speed short-time prediction method based on road section topological structure classification | |
| Foiadelli et al. | Energy consumption prediction of electric vehicles based on big data approach | |
| CN115422747A (en) | Method and device for calculating discharge amount of pollutants in tail gas of motor vehicle | |
| Pan et al. | Development of an energy consumption prediction model for battery electric vehicles in real-world driving: A combined approach of short-trip segment division and deep learning | |
| CN113780665A (en) | Private car stop position prediction method and system based on enhanced recurrent neural network | |
| CN111784027A (en) | Urban range electric vehicle charging demand prediction method considering geographic information | |
| CN113610298B (en) | User travel energy consumption prediction and path recommendation method considering user travel behaviors |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170908 |
|
| RJ01 | Rejection of invention patent application after publication |